// SPDX-FileCopyrightText: Copyright (c) 2022 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
// SPDX-License-Identifier: MIT

syntax = "proto3";

package nvidia.riva.tts;

option cc_enable_arenas = true;
option go_package = "nvidia.com/riva_speech";

import "riva/proto/riva_audio.proto";
import "riva/proto/riva_common.proto";

service RivaSpeechSynthesis {
  // Used to request text-to-speech from the service. Submit a request
  // containing the desired text and configuration, and receive audio bytes in
  // the requested format.
  rpc Synthesize(SynthesizeSpeechRequest) returns (SynthesizeSpeechResponse) {}

  // Used to request text-to-speech returned via stream as it becomes available.
  // Submit a stream of SynthesizeSpeechRequest messages with desired text and
  // configuration, and receive a stream of bytes in the requested format.
  rpc SynthesizeOnline(stream SynthesizeSpeechRequest)
      returns (stream SynthesizeSpeechResponse) {}

  // Enables clients to request the configuration of the current Synthesize
  // service, or a specific model within the service.
  rpc GetRivaSynthesisConfig(RivaSynthesisConfigRequest)
      returns (RivaSynthesisConfigResponse) {}
}

message RivaSynthesisConfigRequest {
  // If model is specified only return config for model, otherwise return all
  // configs.
  string model_name = 1;
}

message RivaSynthesisConfigResponse {
  message Config {
    string model_name = 1;
    map<string, string> parameters = 2;
  }

  repeated Config model_config = 1;
}

// Required for Zero Shot model
message ZeroShotData {
  // Audio prompt for Zero Shot model. Duration should be between 3 to 10
  // seconds.
  bytes audio_prompt = 1;
  // Sample rate for input audio prompt. Current defaults to 22050, can be user
  // provided.
  int32 sample_rate_hz = 2;
  // Encoding of audio prompt. Supported encodings are LINEAR_PCM and OGGOPUS.
  AudioEncoding encoding = 3;
  // The number of times user wants to pass audio through decoder. This ranges
  // between 1-40. Defaults to 20.
  int32 quality = 4;
  // Transcript corresponding to audio_prompt.
  string transcript = 5;
}

message SynthesizeSpeechRequest {
  // Text to be converted to audio
  string text = 1;
  string language_code = 2;
  // audio encoding params
  AudioEncoding encoding = 3;

  //  The sample rate in hertz (Hz) of the audio output requested through
  //  `SynthesizeSpeechRequest` messages. Models produce an output at a fixed
  //  rate. The sample rate enables you to resample the generated audio output
  //  if required. You use the sample rate to up-sample or down-sample the audio
  //  for various scenarios. For example, the sample rate can be set to 8kHz
  //  (kilohertz) if the output audio is desired for a low bandwidth
  //  application. The sample rate values below 8kHz will not produce any
  //  meaningful output. Also, up-sampling too much will increase the size of
  //  the output without improving the output audio quality.

  int32 sample_rate_hz = 4;
  // voice params
  string voice_name = 5;

  // Zero Shot model params
  ZeroShotData zero_shot_data = 6;

  // A string containing comma-separated key-value pairs of
  // grapheme and corresponding phoneme separated by double spaces.
  string custom_dictionary = 7;

  // Free-form key/value parameters for the synthesizer. Lets the server
  // accept additional or model-specific options without proto changes.
  // Currently recognized keys include "exaggeration_factor".
  map<string, string> custom_configuration = 8;

  // The ID to be associated with the request. If provided, this will be
  // returned in the corresponding response.
  RequestId id = 100;
}

message SynthesizeSpeechResponseMetadata {
  // Currently experimental API addition that returns the input text
  // after preprocessing has been completed as well as the predicted
  // duration for each token.
  // Note: this message is subject to future breaking changes, and potential
  // removal.
  string text = 1;
  string processed_text = 2;
  repeated float predicted_durations = 8;
}

message SynthesizeSpeechResponse {
  bytes audio = 1;
  SynthesizeSpeechResponseMetadata meta = 2;

  // The ID associated with the request
  RequestId id = 100;
}